C2W4: Data Augmentation Arguments

In the course notebook:

training_datagen = ImageDataGenerator(
      rescale = 1./255,
      rotation_range=40,
      width_shift_range=0.2,
      height_shift_range=0.2,
      shear_range=0.2,
      zoom_range=0.2,
      horizontal_flip=True,
      fill_mode='nearest')

However, in the rock-paper-scissors data set, shear is bad.
Removing this argument dramatically improves val_loss.

It looks like it could be corrected on that point.

Thanks for sharing the results.